[dpdk-dev] [PATCH] vhost: support UDP Fragmentation Offload

[Hu, Jiayu]

Hi Maxime,

> -----Original Message-----
> From: Maxime Coquelin [mailto:maxime.coquelin at redhat.com]
> Sent: Wednesday, December 6, 2017 4:37 PM
> To: Hu, Jiayu <jiayu.hu at intel.com>; dev at dpdk.org
> Cc: yliu at fridaylinux.org; Tan, Jianfeng <jianfeng.tan at intel.com>
> Subject: Re: [dpdk-dev] [PATCH] vhost: support UDP Fragmentation Offload
> 
> 
> 
> On 11/21/2017 07:56 AM, Jiayu Hu wrote:
> > In virtio, UDP Fragmentation Offload (UFO) includes two parts: host UFO
> > and guest UFO. Guest UFO means the frontend can receive large UDP
> packets,
> > and host UFO means the backend can receive large UDP packets. This patch
> > supports host UFO and guest UFO for vhost-user.
> >
> > Signed-off-by: Jiayu Hu <jiayu.hu at intel.com>
> > ---
> >   lib/librte_mbuf/rte_mbuf.h    |  7 +++++++
> >   lib/librte_vhost/vhost.h      |  2 ++
> >   lib/librte_vhost/virtio_net.c | 10 ++++++++++
> >   3 files changed, 19 insertions(+)
> >
> > diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
> > index ce8a05d..3d8cfc9 100644
> > --- a/lib/librte_mbuf/rte_mbuf.h
> > +++ b/lib/librte_mbuf/rte_mbuf.h
> > @@ -209,6 +209,13 @@ extern "C" {
> >   /* add new TX flags here */
> >
> >   /**
> > + * UDP Fragmentation Offload flag. This flag is used for enabling UDP
> > + * fragmentation in SW or in HW. When use UFO, mbuf->tso_segsz is used
> > + * to store the MSS of UDP fragments.
> > + */
> > +#define PKT_TX_UDP_SEG	(1ULL << 42)
> > +
> > +/**
> >    * Request security offload processing on the TX packet.
> >    */
> >   #define PKT_TX_SEC_OFFLOAD 		(1ULL << 43)
> > diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
> > index 1cc81c1..fc109ef 100644
> > --- a/lib/librte_vhost/vhost.h
> > +++ b/lib/librte_vhost/vhost.h
> > @@ -206,10 +206,12 @@ struct vhost_msg {
> >   				(1ULL <<
> VHOST_USER_F_PROTOCOL_FEATURES) | \
> >   				(1ULL << VIRTIO_NET_F_HOST_TSO4) | \
> >   				(1ULL << VIRTIO_NET_F_HOST_TSO6) | \
> > +				(1ULL << VIRTIO_NET_F_HOST_UFO) | \
> >   				(1ULL << VIRTIO_NET_F_CSUM)    | \
> >   				(1ULL << VIRTIO_NET_F_GUEST_CSUM) | \
> >   				(1ULL << VIRTIO_NET_F_GUEST_TSO4) | \
> >   				(1ULL << VIRTIO_NET_F_GUEST_TSO6) | \
> > +				(1ULL << VIRTIO_NET_F_GUEST_UFO) | \
> 
> I actually have the same question as for GSO.
> Dos it impact performance as it seems enabled by default in QEMU?

When enable host/guest UFO, the frontend and the backend can communicate
via large UDP packets, thus reducing the number of packets to be processed.
It's similar with TSO for virtio. Therefore, I think host/guest UFO can bring
performance gains.

> How do you test it?

We launch testpmd with two vhost-user ports, which connect to two VMs in one
server. When launch qemu, we enable host and guest UFO with command "host_ufo=on,
guest_ufo=on,csum=on". Then we run iperf in the two VMs to send large UDP packets.
If you use "show port xstats all" in testpmd, you can see the vhost-user port can receive
large UDP packets, which means host_ufo works. Additionally, if the forwarded large UDP
packets can be received by the VM, it means guest_ufo also works.

Thanks,
Jiayu
> 
> Thanks,
> Maxime
> 
> >   				(1ULL << VIRTIO_RING_F_INDIRECT_DESC)
> | \
> >   				(1ULL << VIRTIO_NET_F_MTU) | \
> >   				(1ULL << VIRTIO_F_IOMMU_PLATFORM))
> > diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
> > index 6fee16e..3a3a0ad 100644
> > --- a/lib/librte_vhost/virtio_net.c
> > +++ b/lib/librte_vhost/virtio_net.c
> > @@ -188,6 +188,11 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf,
> struct virtio_net_hdr *net_hdr)
> >   		net_hdr->gso_size = m_buf->tso_segsz;
> >   		net_hdr->hdr_len = m_buf->l2_len + m_buf->l3_len
> >   					+ m_buf->l4_len;
> > +	} else if (m_buf->ol_flags & PKT_TX_UDP_SEG) {
> > +		net_hdr->gso_type = VIRTIO_NET_HDR_GSO_UDP;
> > +		net_hdr->gso_size = m_buf->tso_segsz;
> > +		net_hdr->hdr_len = m_buf->l2_len + m_buf->l3_len +
> > +			m_buf->l4_len;
> >   	} else {
> >   		ASSIGN_UNLESS_EQUAL(net_hdr->gso_type, 0);
> >   		ASSIGN_UNLESS_EQUAL(net_hdr->gso_size, 0);
> > @@ -834,6 +839,11 @@ vhost_dequeue_offload(struct virtio_net_hdr
> *hdr, struct rte_mbuf *m)
> >   			m->tso_segsz = hdr->gso_size;
> >   			m->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
> >   			break;
> > +		case VIRTIO_NET_HDR_GSO_UDP:
> > +			m->ol_flags |= PKT_TX_UDP_SEG;
> > +			m->tso_segsz = hdr->gso_size;
> > +			m->l4_len = sizeof(struct udp_hdr);
> > +			break;
> >   		default:
> >   			RTE_LOG(WARNING, VHOST_DATA,
> >   				"unsupported gso type %u.\n", hdr-
> >gso_type);
> >